WHAMY is a novel actin polymerase promoting myoblast fusion, macrophage cell motility and sensory organ development in Drosophila

The actin cytoskeleton plays a central role in a number of different cellular functions, such as cell shape changes, cell motility and membrane trafficking. Members of the Wiskott–Aldrich syndrome protein (WASP) family are conserved nucleation-promoting factors (NPF) that activate the Arp2/3 complex, a major actin nucleator in eukaryotic cells (Pollard, 2007; Pollard and Beltzner, 2002; Pollitt and Insall, 2009; Rottner and Stradal, 2011; Takenawa and Suetsugu, 2007). In mammals, the WASP protein family consists of eight different members: the two Wiskott-Aldrich syndrome proteins WASP and N-WASP (also known as WAS and WASL, respectively) (Derry et al., 1994; Miki et al., 1996), the related WASP family Verprolin homologous proteins WAVE1–WAVE3 (also known as SCAR1–SCAR3 and WASF1–WASF3) (Miki et al., 1998), the Wiskott–Aldrich syndrome protein and SCAR homolog WASH (also known as WASH1) (Linardopoulou et al., 2007) and the WHAMM and JMY proteins (Campellone et al., 2008; Zuchero et al., 2009). WASP proteins share a conserved C-terminal Arp2/3-complex-activating WCA module. This module consists of either one or multiple actin-monomer-binding WH2 (W) domains, a central domain (C) and an acidic (A) domain, which mediate Arp2/3 binding (Campellone and Welch, 2010; Stradal and Scita, 2006). Apart from the catalytic WCA module, WASP proteins often share a proline-rich region and a basic region, which bind SH3-domain containing proteins and acidic phosphoinositides, respectively (Suetsugu et al., 2003; Takenawa and Suetsugu, 2007). WASP proteins are regulated by similar molecular principles. Under resting conditions NPFs are primarily inactive and become activated upon binding of the Rho GTPases Cdc42 and Rac1. Additionally, a variety of factors further modulate proper activation and recruitment of WASP proteins (Krause and Gautreau, 2014).

In Drosophila, only three WASP subfamily members have been described, namely WAVE, WASP and WASH (also known as CG13176). Insects like Drosophila have subsequently lost a WHAMM/JMY gene, although the common ancestor first arose in invertebrates (Kollmar et al., 2012; Veltman and Insall, 2010). Genetic studies indicate that WAVE and WASP are the central activators of the Arp2/3 complex, differentially regulating most aspects of Arp2/3 function in Drosophila (Ben-Yaacov et al., 2001; Bogdan et al., 2004,, 2005; Bogdan and Klämbt, 2003; Fricke et al., 2009; Leibfried et al., 2008; Stephan et al., 2011; Zallen et al., 2002; Zobel and Bogdan, 2013). These studies highlight distinct, but also overlapping cellular requirements of WAVE and WASP during development. WAVE function is in particular essential for cell shape and morphogenetic cell movements during development (Stephan et al., 2011; Zallen et al., 2002). By contrast, WASP function is needed for cell fate specification of sensory organ precursors (SOPs) and spermatid maturation (Ben-Yaacov et al., 2001; Berger et al., 2008; Mukherjee et al., 2011; Rotkopf et al., 2011). Both, WASP and WAVE are required for myoblast fusion (Abmayr and Pavlath, 2012; Gildor et al., 2009; Berger et al., 2008).

Loss of maternal and zygotic WASP results in late-embryonic lethality due to strong defects in cell fate decisions of neuronal cell lineages and myoblast fusion defects (Ben-Yaacov et al., 2001; Mukherjee et al., 2011; Schafer et al., 2007; Massarwa et al., 2007). Remarkably, animals lacking zygotic WASP function survive until early adulthood (Ben-Yaacov et al., 2001). Thus, maternally provided WASP protein is sufficient for proper embryonic and larval development (Ben-Yaacov et al., 2001). Mutant wasp flies show no strong morphological defects except a partial loss of sensory bristles. Loss of zygotic Arp2/3 function results in a similar, albeit stronger, neurogenic phenotype suggesting an involvement of additional factors in Arp2/3-dependent SOP development (Rajan et al., 2009). The loss of sensory bristles in wasp and arp2/3 mutants phenocopies Notch loss-of-function and is caused by a pIIa-to-pIIb cell fate transformation (Ben-Yaacov et al., 2001). This results in an excess of neurons at the expense of bristle sheath, shaft and socket cells (Ben-Yaacov et al., 2001; Rajan et al., 2009). Recent work further suggests that the WASP–Arp2/3 pathway rather plays an important role in the trafficking of Delta-positive vesicles from the basal area to the apical cortex of the signal-sending pIIb cell (Rajan et al., 2009).

Remarkably, rescue experiments have implied that established activators of WASP, such as Cdc42 or phosphatidylinositol 4,5-bisphosphate (PIP₂), are not required for WASP function, neither for the myoblast fusion process nor for SOP development (Ben-Yaacov et al., 2001). The identity of an independent activator that might act cooperatively to control Arp2/3 function in these contexts is unknown. Here, we present a functional analysis of WHAMY, a new WASP-like protein that regulates cell motility of Drosophila blood cells but also synergizes with WASP during embryonic muscle formation and cell fate specification of adult SOPs.

Sequence analysis revealed that WHAMY is a paralog of WASP, which arose from duplication in the genus Drosophila. WASP and WHAMY carry two WH2 domains in their C-terminus (Fig. 1A). Although WASP contains one CRIB domain (amino acids 230–293) and an N-terminal WH1 domain (amino acids 31–140), WHAMY has two N-terminal CRIB domains (amino acids 4–70 and 185–254; Fig. 1A). As the domain architecture of WASP is conserved throughout the eukaryotes, this indicates a loss of the WH1 domain in WHAMY followed by an accretion of a CRIB domain. To unravel the origin of this domain, a phylogenetic tree of CRIB domains of both proteins from different Drosophila species and, as an outgroup, Anopheles gambiae and Culex quinquefasciatus, was calculated (Fig. 1B; alignment in Table S1). Here, the Drosophila WASP CRIB domains were the clear outgroup to both WHAMY domains. Thus, an internal duplication is the origin of the second CRIB domain in WHAMY.

Gene duplication frequently leads to a phase of relaxed selection (Lynch and Conery, 2000). This can be followed by neofunctionalization, indicated by a phase of positive selection. In addition, in the case of the CRIB duplication, models allowing for a different (faster) evolution rate after duplication fitted the data significantly (P<0.01) better than those assuming a single rate over the whole tree (Yang, 2007). More detailed models assuming an increased rate after duplication of WASP and duplication of the CRIB domains fitted the data as good as models assuming an increased rate over the whole WHAMY subtree (Fig. S1). This increased rate in WHAMY compared to WASP was also found when analysing the region C-terminal to the CRIB domains independently (Table S2). Thus, the duplication of the whole gene had a comparable effect on the WHAMY CRIB domains to that of their succeeding duplication. This could indicate that the neofunctionalization of the CRIB domains is still in progress. Still, even considering site-specific models, no substantial traces of a phase of positive selection were detected (Yang, 1998). A multiple sequence alignment of WASP and WHAMY CRIB domains allowed for the identification of sites that might be involved in a change of function (Fig. 1C; Table S1). Most interestingly, there were sites that showed a different conservation between WASP and WHAMY. These differences affected one or both WHAMY CRIB domains. Mapping these sites onto the structure of a WASP CRIB domain (PDB ID 1CEE) together with the ligand shows that these sites are in positions responsible for the specificity of ligand binding (Abdul-Manan et al., 1999; Fig. 1C). These sites might therefore be involved in a change of substrate specificity for the CRIB domains in WHAMY.

Based on the above prediction, we first tested whether the two CRIB domains of WHAMY are functional binding sites for active Cdc42 or Rac1. We performed pulldown experiments using either GST–Cdc42 or GST–Rac1 incubated with lysates prepared from S2R+ cells expressing different WHAMY–EGFP fusions lacking either one or both CRIB domains. Full-length WHAMY–EGFP strongly bound Cdc42, but this binding was, however, independent of the nucleotide-binding status of Cdc42 (Fig. 1D). The truncated WHAMY proteins lacking either the first (ΔCRIB1) or the second CRIB domain (ΔCRIB2) still bound Cdc42-GDP. However, both CRIB domains differed in their capacity to bind to activated Cdc42. Only the deletion of the second CRIB domain led to loss of binding. Thus, the differences observed in the sequence between the two CRIB domains indeed result in functional divergence. Notably, WHAMY interacted specifically with GTP-loaded Rac1. Pulldown experiments with ΔCRIB1 or ΔCRIB2 deletions further indicated that, in contrast to with Cdc42, both CRIB domains of WHAMY were equally required for the interaction with activated Rac1. Again, loss of both CRIB domains entirely abolished Rac1 binding (Fig. 1D). Therefore, we conclude that, unlike WASP, WHAMY is not an effector of Cdc42 but rather acts downstream of Rac1.

We next examined the subcellular localization of WHAMY–EGFP in transfected S2R+ cells. WHAMY–EGFP strongly localized at the leading edge and at dynamic vesicles (Fig. 1E,E′; Movie 1). Interestingly, deletion of both CRIB domains prevented WHAMY targeting to lamellipodial tips, whereas the vesicular localization was not affected (Fig. 1F,F′; Movie 2). Thus, we conclude that Rac1 binding mediates the localization of WHAMY to lamellipodial tips.

Structured illumination microscopy (SIM) further revealed that WHAMY-marked vesicles overlapped with F-actin, suggesting that the vesicles move at the tips of actin tails (Fig. 1G; see also Movies 1 and 2). Co-transfection experiments with different Rab–Cherry fusions indicated that a large majority of the vesicles marked by WHAMY–EGFP were Rab11-positive recycling endosomes (mean Pearson coefficient, 0.67; Fig. 1H,I; Movie 3).

The prominent localization of WHAMY at the leading edge prompted us to analyze the activity of WHAMY in regulating actin assembly in vitro. Recombinant full-length WHAMY (WHAMY-FL) was sufficient to stimulate actin polymerization in a concentration-dependent manner, albeit with markedly slower kinetics as compared to the C-terminal half of the Drosophila formin Diaphanous (Dia; Fig. 2A). Thus, we reasoned that WHAMY-FL might be autoinhibited and require interaction with the small GTPase Rac1 to display full activity. Of note, in pulldown experiments active GST–Cdc42 showed stronger interaction with WHAMY-FL as compared to active GST–Rac1 (Fig. S2A). However, given that the Cdc42–WHAMY interaction was unspecific (Fig. 1D), we continued the in vitro analysis on WHAMY-FL with constitutively active Rac1-V12. Increasing concentrations of Rac1 further stimulated WHAMY-FL-mediated actin polymerization, indicative of WHAMY activation by this GTPase (Fig. 2B).

Given that other members of the Drosophila WASP protein family act as NPFs for the Arp2/3 complex, we next tested whether WHAMY additionally exhibits NPF activity towards the Arp2/3 complex. However, and even in the presence of active Rac1, WHAMY-FL did not activate the Arp2/3 complex (Fig. 2C). By contrast, Cdc42-activated WASP promoted Arp2/3-complex-mediated actin assembly even at low nanomolar concentrations (Fig. 2C). To explore the potential synergy between WHAMY and WASP observed during morphogenesis (see below), we performed GST pulldown assays with recombinant GST–WASP and WHAMY-FL alone and in presence of constitutively active Cdc42 and Rac1. Indeed, a substantial fraction of recombinant WHAMY-FL was pulled down by GST–WASP-coated beads but not by the control beads coated with GST alone (Fig. 2D). Surprisingly, the WHAMY–WASP interaction was not further strengthened by addition of Cdc42 or Rac1. Control experiments without WASP confirmed that activated Cdc42 and activated Rac1 alone did not affect the amount of WHAMY pulled down with GST-coated beads (Fig. S4B).

Up to an equimolar ratio, WHAMY-FL-mediated actin polymerization was stimulated by increasing concentrations of Cdc42-activated WASP, and this effect was augmented even more by active Rac1 (Fig. 2E). Addition of WASP at concentrations above 200 nM, however, resulted in decreased actin polymerization rates and lower plateau values indicative of actin monomer sequestration by WASP WH2 domains. Next, we assayed Arp2/3 complex activation by WASP in the presence of WHAMY-FL. Although WHAMY-FL did not directly stimulate the ability of WASP to activate the Arp2/3 complex, it enhanced the Arp2/3-complex-dependent actin polymerization mediated by Cdc42-activated WASP (Fig. 2F). Consistent with previous studies, we also found that full-length WASP alone was not entirely autoinhibited with regard to Arp2/3 complex activation (Fricke et al., 2009; Rohatgi et al., 1999). Given the observed WHAMY–WASP interaction, these findings therefore suggest that WHAMY might generate mother filaments in close vicinity of the Arp2/3 complex to promote Arp2/3-complex-dependent actin assembly.

To gain deeper insights into the activities of WHAMY, we performed single-filament total internal reflection fluorescence (TIRF) microscopy assays. In the presence of 100 nM WHAMY-FL, about three times more filaments were formed as compared to the actin control, indicating moderate nucleation activity of WHAMY-FL (Fig. 3A,B). Notably, WHAMY formed bright actin-rich clusters at the sides or the ends of the filaments that were capable of nucleating new actin filaments in the absence of profilin (Movie 4). Intriguingly, upon binding of free filament barbed ends to WHAMY clusters, these filaments were rapidly elongated in short bursts with up to >200 subunits/s demonstrating exceptional actin polymerase activity (Fig. 3C–E; Movie 4). In the presence of 5 µM profilin, the nucleation activity of WHAMY was almost abolished and the filament elongation rate diminished to ∼90 subunits/s. Interestingly, the WHAMY clusters as well as WHAMY-elongated filaments appeared much brighter in these conditions (Fig. 3D; Movie 4). Given that profilin exhibits low affinity to Cys³⁷⁴-labeled actin (Kovar et al., 2006), these data indicate that WHAMY cannot efficiently bind to profilin–actin and instead mainly recruits uncomplexed G-actin, which is predominantly the labeled species at the five-fold molar excess of profilin used in this assay. Consistently, after detachment from the WHAMY clusters, these filaments grew with the same speed and brightness as the control filaments. To gain more insight into a possible oligomerization of WHAMY, we performed analytical size exclusion chromatography with purified WHAMY-FL, and obtained a calculated molecular mass of 413 kDa, strongly suggesting oligomerization of the 64-kDa protein, although a precise oligomerization state cannot be calculated from these experiments without knowledge about the shape of the molecule (Fig. S3A,B). Combined, these data reveal only moderate nucleation, but exceptionally fast filament elongation activity of clustered WHAMY, which could be utilized to generate mother filaments for the Arp2/3 complex. In this case, they also support the concept of an inhibitory role of profilin in Arp2/3-complex-mediated actin assembly (Suarez et al., 2015).

Next, we analyzed the in vivo function of whamy. The whamy gene is located on the third chromosome at the cytological location 85E8 (Fig. 4A). It consists of four exons encoding three predicted protein isoforms that differ only in their N-termini (FlyBase). The largest protein isoform, WHAMY^PC (66 kDa) contains both N-terminal CRIB domains, whereas the shortest isoform WHAMY^PB (54.1 kDa) lacks the first CRIB domain (FlyBase). To elucidate the in vivo role of WHAMY, we generated mutant flies by imprecise excision mutagenesis utilizing two different P-elements inserted in the 3′-UTR of the whamy gene (Fig. 4A).

We isolated two deletions, whamy^Δ109 and whamy^Δ65, removing either the third and fourth exon as well as parts of the second exon. Reverse transcriptase (RT)-PCR and western blot analyses confirmed a loss of WHAMY expression in whamyΔ109 and whamyΔ65 deletion mutants (Fig. 4B; Fig. S3C). In wild-type extracts, a distinct WHAMY-specific protein band was detected at ∼60 kDa (Fig. 4B).

Remarkably, flies with all whamy alleles in homozygosity and trans-heterozygosity were viable and fertile with no obvious morphological defects. Hence, we conclude that WHAMY is not essential for Drosophila development, but rather might regulate cell-type-specific actin-dependent processes. High-throughput expression data indicate that WHAMY is most abundant in the larval and adult gut, and more weakly expressed in Malpighian tubules and fat body (FlyAtlas, anatomical expression database), but low expression is also found in hemocyte-like S2 cells and tumorous blood cells mbn2 (modENCODE cell line expression database). Therefore, we further analyzed the cellular WHAMY function in larval and pupal blood cells, an excellent model system for studying functions of genes involved in actin dynamics at a single-cell level (Sander et al., 2013).

Isolated macrophage-like blood cells showed a broad lamellipodial actin filament network when spread on cover glasses (Fig. 4C). In contrast, whamy-deficient macrophages displayed strong defects in cell flattening and cell spreading resulting in significantly smaller cells (Fig. 4D,E, quantification in Fig. 4F). Re-expression of a WHAMY–EGFP transgene in homozygous whamy^Δ109 mutant macrophages rescued the cell-spreading defects, indicating that the defects resulted from a loss of WHAMY function (Fig. 4G). Similar to transfected S2R+ cells, WHAMY–EGFP localized at the leading edge and at actin-rich vesicles in isolated macrophages (Fig. 4H′,H″). Remarkably, rescued cells even showed an increased cell spreading behavior, indicating that WHAMY promotes cell spreading (Fig. 4G,H). Increasing the levels of WHAMY protein by transgene overexpression in a wild-type background did not further increase cell size; rather, overexpression of WHAMY in fact resulted in a marked reduction of cell size and instead dramatically induces the formation of numerous filopodial protrusions (Fig. 4I, quantification in Fig. 4J,K; Movie 5). Overexpressed WHAMY–EGFP still localized to the lamellipodium and vesicles, and strongly marked the tips of induced filopodia (Fig. 4I–I‴; Movie 5). Elevated levels of WHAMY in a wild-type background also induced a dramatic increase of filopodial protrusions in vivo (Movie 6). Thus, we propose that WHAMY-mediated actin assembly is required for cell spreading and membrane protrusions of Drosophila blood cells. Given that wasp mutant cells showed no defects in cell spreading or membrane protrusions (Fig. 4F), WHAMY seems to have acquired a new function in cell motility during evolution.

We next dissected the function of WHAMY in migrating macrophages at different pupal stages in vivo. Wild-type prepupal macrophages acquired a spread morphology with broad, polarized lamellipodia (Fig. 5A; Movie 7). First, we examined the localization of the WHAMY–EGFP transgene in homozygous whamy^Δ109 mutant cells lacking the endogenous protein. The WHAMY–EGFP transgene efficiently rescued not only the cell spreading defects of isolated mutant cells, but also did not induce the formation of ectopic filopodial protrusions (Fig. 5B; Movies 8 and 9). WHAMY strongly localizes at the leading edge of migrating polarized macrophages (arrows in Fig. 5B, Movies 8 and 9) and in addition marks dynamic vesicular structures in the cell body of migrating cells (arrowheads in Fig. 5B, Movies 8 and 9).

Given the distinct localization of WHAMY at the leading edge of the cell, we further examined a possible role in cell migration. As recently reported, wild-type macrophages started to initiate random single-cell migration 2 h after puparium formation (Moreira et al., 2013) (Fig. 5C; Movie 10). Compared to wild-type cells, whamy^Δ109 mutant cells showed an impaired spread morphology and had a reduced cell size (Fig. 5D, quantification in Fig. 5E; Movie 11). Striking differences were also seen in the migratory behavior of mutant cells. Whereas wild-type cells spread out and dispersed at 2 h after puparium formation (APF), whamy^Δ109 mutant macrophages still remained closely clustered together and formed densely packed groups of single cells (Fig. 5D, quantification in Fig. 5F; Movie 11). Tracking of individual migrating cells revealed a significant reduction in cell speed (Fig. 5G). Control cells migrated with an average maximum speed of 3.3±0.9 µm/min (mean±s.d.), whereas whamy^Δ109 mutant cells migrated considerably slower with an average maximum speed of 1.9±0.7 µm/min (Fig. 5G). This result indicates that WHAMY is required for proper cell spreading and cell migration in vivo.

To further elucidate the importance of WHAMY in directed cell migration, we next performed in vivo wounding experiments. Macrophages migrate towards a laser-induced wound or after ablation of a single cell (the ablation side is marked by a yellow circle in Fig. 5H,H′; Movie 12). Similar to control cells, whamy-deficient cells still polarized in response to wounds and exhibited a directional cell migratory behavior, albeit with significantly reduced speed (Fig. 5I,I′; Movie 13). To quantify changes in directional cell migration, we analyzed the migratory behavior of macrophages in later pupal stages (17 h APF; Fig. 5I,J) by measuring the Histogram-based Macrophage Migration Score (HMMS) that we recently established (Lammel et al., 2014). Macrophages from later pupal stages are more dispersed, migrating in 3D rather than in 2D, as in prepupae, and their wound response is much stronger (Sander et al., 2013; Lammel et al., 2014). HMMS measurements of wild-type and mutant cells confirmed that cells deficient for whamy were still able to respond and migrate to wound sites but that their speed was reduced (Fig. 5H–J; Movies 14 and 15). The differences between wild-type and mutant macrophages were most obvious in the first 600 s after wounding, when a rapid response of the wild-type cells compared to whamy mutant cells could be measured (Fig. 5J). These data indicate that WHAMY is required for proper cell spreading and cell migration in vivo.

Given the observed in vitro WHAMY–WASP interaction, we next conducted co-overexpression experiments in vivo using the Gal4-UAS system (Brand and Perrimon, 1993). Forced expression of full-length WASP in wing imaginal disc epithelium did not induce actin assembly, suggesting that WASP is strongly auto-inhibited in vivo (Fig. 6B). WHAMY overexpression only slightly increased F-actin formation (Fig. 6C). However, co-overexpression of WHAMY and WASP strongly induced F-actin (Fig. 6D), suggesting that there are synergistic functions of both proteins in vivo. How might WHAMY and WASP act together in vivo? In S2R+ cells, we found not only an interaction between a WHAMY–EGFP and a WASP–Cherry fusion but also a striking colocalization at vesicles (Fig. S4A,B). However, there was no overlap at the leading edge of S2R+ cells, suggesting a common function of WHAMY and WASP in vesicle trafficking (Fig. S4A).

To further test functional interactions in vivo, we analyzed trans-heterozygous and homozygous double mutants. For this analysis we used the wsp¹ allele, which contains a small intragenic deletion resulting in a premature stop codon (Ben-Yaacov et al., 2001). As for all other EMS-induced wasp alleles, the chromosome carrying the wasp¹ allele also contains additional lethality-causing mutations. Therefore, we used transheterozygous combinations with the deficiency Df(3R)3450, which completely removes, among others, the wasp gene (Ben-Yaacov et al., 2001). wsp¹/Df(3R)3450 transheterozgygotes survived until early adulthood, but many mutant flies failed to fully eclose from the pupal case. Escaper flies exhibited a lack of sensory microchaetes and macrochaetes (Ben-Yaacov et al., 2001). To analyze wasp, whamy double mutant flies, we cleaned the wsp¹ allele from associated lethal mutations (designated as wsp^1*) and generated recombinant flies carrying the wsp¹ and the whamy^Δ109 allele (wsp^1*, whamy^Δ109). Homozygous wsp^1* flies survived as wsp¹/Df(3R)3450 transheterozgygotes until early adulthood with no significant phenotypic differences (Fig. 6H). However, homozygous wsp^1*, whamy double mutants were late-embryonic lethal with very few larval escapers (Fig. 6H). Even the loss of one whamy gene copy in wsp^1*/Df(3R)3450 transheterozgygotes (wsp^1*, whamy^Δ109/Df(3R)3450) significantly shifted the lethality from late pupal or adult to early larval stages (Fig. 6H). These animals survived until late pupal stage (pharate adults) but all failed to eclose from the pupal case and show an enhanced loss of sensory bristles phenotype compared to homozygous or hemizygous wsp^1* mutants (compare Fig. 6E–G, quantification of the increased lethality in Fig. 6H). Quantification of this neurogenic phenotype revealed that the formation of the sensory bristles was strongly affected, similarly to what is observed for arp2/3 mosaics (Fig. 6I; Rajan et al., 2009).

Double staining of pupal nota dissected at 24 h APF with the neuronal nuclear marker Elav, and Cut, which mark all sensory organ precursor cells, confirmed that the enhanced loss of external sensory bristle structures was due to an increased excess of neurons (Fig. 6J). In wild-type flies, each sensory organ precursor cell (SOP, pI cell) normally undergoes an asymmetric cell division to generate two different daughter cells (pIIa and pIIb). The pIIa cell divides to generate the external sensory organ cells, the shaft and socket cells, whereas the pIIb cell undergoes two additional asymmetric cell divisions to generate the internal sensory organ cells including the neuron, the sheath cell and a glial cell (Gho et al., 1999). WASP–Arp2/3-mediated actin nucleation seems to be required for both cell fate decisions. wasp mutants frequently showed a pIIa-to-pIIb cell fate transformation resulting in a duplication of neurons (Ben-Yaacov et al., 2001; Fig. 6J). By contrast, trans-heterozygous wasp, whamy double mutants had, on average, three Elav-positive neurons per sensory organ (Fig. 6J,K). Thus, WHAMY and WASP function cooperatively during sensory organ development. Given the observed synergistic effects of WHAMY and WASP on actin assembly, we further tested whether WHAMY expression could complement wasp function. We found that forced expression of WHAMY in wasp mutant sensory clusters using the scabrous Gal4 driver resulted in a weak but significant increase of sensory bristles (Fig. 6L). Therefore, WHAMY can partially substitute for WASP functions in sensory organ development.

The strong synergistic defects of trans-heterozygous mutants raise the issue of whether the late-embryonic lethality of whamy, wasp double mutants is caused by cell fate decision defects during sensory organ development as previously shown for embryos lacking both maternal and zygotic WASP (Ben-Yaacov et al., 2001). However, we found no significant changes in the number of neurons in the mature embryonic PNS of whamy, wasp double mutants. Given that WASP function is further required for multinucleated muscle formation during embryogenesis and for the attachment of muscles to epidermal tendon cells (Massarwa et al., 2007; Schafer et al., 2007), we further analyzed possible defects during muscle development. Remarkably, single whamy and wasp mutants showed similar attachment defects of the longitudinal transverse muscles (Fig. 7B,C, arrows). By contrast, whamy, wasp double mutants displayed strong defects in muscle size and morphology (Fig. 7D). Multinucleated muscle formation depends on the fusion of two myoblast populations: founder cells and fusion-competent myoblasts (FCMs). Founder cells initiate muscle formation by attracting FCMs for the fusion process (Bate, 1993). They further determine the position, size, orientation, epidermal attachment site of the muscle and patterns of innervation (Bate, 1993). The selection of founder cells depends on Notch-mediated lateral inhibition (Carmena et al., 1995). To further characterize the double mutant phenotype in more detailed, we crossed the rP298-lacZ enhancer trap line that marks the nuclei of founder cells into the background of whamy, wasp double mutants (Nose et al., 1998). In stage 16 embryos, we found that the segmental border muscle contained only one to two nuclei (Fig. 7E, arrows), instead of up to seven nuclei, like in wild-type embryos (Bataille et al., 2010). This finding indicates that myoblast fusion is disturbed in whamy, wasp double mutants. To rule out that this phenotype is based on founder cell specification defects, we compared rP298-lacZ expression at stage 11 embryos of wild-type and whamy, wasp double mutants when myoblasts start to fuse. In both, we observed a comparable number of rP298-lacZ-expressing cells, suggesting that the majority of founder cells are specified correctly (Fig. 7F,G). To determine whether WHAMY localizes, like WASP, at the contact site of adhering myoblasts (Schafer et al., 2007; Deng et al., 2015), we expressed the WHAMY–EGFP transgene in all myoblasts by using the Mef2-GAL4 driver line. We indeed could detect WHAMY–EGFP at the myoblast membrane, but failed to observe a distinct localization of WHAMY at the contact site of myoblasts. Instead, we noticed the expression of WHAMY–EGFP at the attachment sites (Fig. 7H,I, asterisks). A similar distribution has been reported for WASP (Schafer et al., 2007). Interestingly, the localization of WHAMY to muscle attachment sites was lost when a WHAMY deletion that lacked both CRIB domains was expressed in all myoblasts (Fig. 7J). Thus, both CRIB domains seem to be essential for the localization of WHAMY to attachment sites, but this does not seem to disturb muscle attachment (Fig. 7J). Taken together, these data indicate that WASP and WHAMY synergistically function in embryonic myoblast fusion, muscle attachment and adult sensory organ development.

The identification of all WASP family homologs in all sequenced organisms allows a detailed phylogenetic analysis of the origin of diverse subfamilies evolving differential cellular functions. WASP proteins are multi-domain proteins. They share functions that are encoded by similar domains at the C-termini, whereas different N-terminal domains mainly define their diverse cellular processes. Gene duplication and domain shuffling are two important mechanisms driving novel and increasingly complex developmental programs during evolution (Babushok et al., 2007; Buljan and Bateman, 2009). It is thought that this boost in domain shuffling is responsible for the apparent disconnection between greatly increased phenotypic complexity and a relatively small difference in gene number between humans and Drosophila (Babushok et al., 2007; Claverie, 2001).

The whamy gene is an excellent example for how gene duplication and subsequent domain shuffling can create new gene functions after initial gene duplication. It arose through a duplication of wasp at the base of the genus Drosophila. Although the encoded protein has evolved a new function in cell motility, it also functions synergistically with WASP in muscle formation and sensory organ development. In the latter, WHAMY can even partially substitute for WASP, indicating that it has kept functionality following the duplication. This duality is reflected in the sequence of the WHAMY CRIB domains. As there is an overlap in function with WASP, selective pressure has been reduced since the duplication, leading to the observed increase of evolutionary rate. Following the duplication of the CRIB domains within WHAMY, a similar trend can be found. Whereas one domain has kept the function of binding to Cdc42-GTP, the other has lost the ability to interact. This is reflected in domain-specific conserved substitutions. The duplication of the wasp gene and subsequent subneofunctionalization of whamy might have occurred at the same time as the loss of a true WHAMM/JMY ancestor during insect evolution (Veltman and Insall, 2010). Like Drosophila WHAMY, the common ancestor of WHAMM/JMY proteins in invertebrates also lacks the characteristic C-terminal tryptophan residue in their VCA domains that is crucial for Arp2/3 binding and activation (Veltman and Insall, 2010). This further implies a primary Arp2/3-independent function of the common ancestor of invertebrate WHAMM/JMY proteins.

WHAMY shows no Arp2/3-activating NPF activity in vitro. However, different from WASP, WHAMY itself is able to promote fast elongation of linear actin filaments from actin-rich clusters. With respect to its activity, WHAMY resembles the WH2-domain containing Ena/VASP polymerases that actively drive processive actin-filament elongation and promote assembly of both lamellipodial and filopodia actin networks (Breitsprecher et al., 2008,, 2011; Hansen and Mullins, 2010). Notably, Ena/VASP proteins are tetramers, and their oligomerization is mandatory to allow for polymerase activity in experiments in solution, as used in this study (Breitsprecher et al., 2008). As we have exclusively observed fast filament elongation from WHAMY clusters in our TIRF experiments, and consistent with the size exclusion chromatography experiments, we propose that WHAMY requires oligomerization to acquire actin polymerase activity. Concerning previously analyzed proteins of the WASP family, the filament elongation activity of WHAMY is therefore rather unique, and when compared to other fast actin polymerases, only the Drosophila formin Diaphanous achieves comparable high elongation activity in vitro (Yan et al., 2013). As evidenced from the pyrene data, the activity of WHAMY can further be increased by Rac1.

Rac1 seems to act on both the activity and the localization of WHAMY at lamellipodial tips. Both of the two CRIB domains of WHAMY bind equally to activated Rac1, and only loss of both CRIB domains abolishes Rac1 binding and the localization to the leading edge. Therefore, it currently remains unclear why WHAMY contains two CRIB domains and whether they differentially mediate distinct cellular functions. They might contribute to a local clustering of WHAMY and Rac1 at the leading edge. The most prominent Rac1 effector represents the WAVE regulatory complex (WRC) that drives Arp2/3-mediated branched actin nucleation. Rac1 directly binds and activates the WRC by allosterically releasing the bound Arp2/3-activating WCA domain of WAVE (Chen et al., 2010; Eden et al., 2002). Overexpression of WHAMY leads to a strong induction of filopodia, presumably due to the filament elongation activity of WHAMY. Additionally, competition between WHAMY and the WRC for Rac1 could disturb the balance between nucleation and elongation activity, and therefore might contribute to the observed overexpression phenotype. Different from WHAMY, WRC function is essential for lamellipodia formation and cell migration in most eukaryotic cells (Pollitt and Insall, 2009). By contrast, loss of WHAMY function does not impair lamellipodia formation but rather regulates cell spreading and contributes to cell motility.

WHAMY does not compete but rather functions together with WASP in Drosophila morphogenesis. Previous studies have revealed that the major established activators of WASP, such as Cdc42 and PIP₂, are not required for the function of WASP in sensory organ development or myoblast fusion (Tal et al., 2002). This observation already suggests that additional components, such as WHAMY, might act together with WASP in sensory organ development and myoblast fusion. Consistent with this, we found that further reduction of whamy function in wasp mutants phenocopies loss of arp2/3 function, resulting in an excess of neurons and a near absence of bristle sheath, shaft and socket cells (Ben-Yaacov et al., 2001; Rajan et al., 2009). Our rescue data further indicate that WHAMY can partially substitute for WASP function. Thus, WHAMY cooperates with WASP rather than acting redundantly in sensory organ development. Based on our TIRF microscopy data, we suggest that WHAMY might potentially generate mother filaments in close vicinity of Arp2/3 complex facilitating Arp2/3-mediated actin assembly.

How might WHAMY and WASP act on actin dynamics during sensory organ development? Recent work suggests that the WASP–Arp2/3 pathway is not involved in Notch receptor endocytosis or its processing in the signal-receiving cell (pIIa) but rather plays an important role in the trafficking of Delta-positive vesicles from the basal area to the apical cortex of the signal-sending pIIb cell (Rajan et al., 2009). This model also implies that recycled Notch ligands such as Delta and Serrate are active at apical junctions with actin-rich structures induced by WASP and the Arp2/3 complex (Rajan et al., 2009), which in turn activate apical Notch receptor in pIIa. In vivo, WHAMY localizes at dynamic vesicles during sensory organ precursor formation (Movie 16) and, together with WASP, becomes strongly enriched at apical junctions shortly after SOP division (Movie 17). Thus, we propose a scenario in which WASP and WHAMY might act either on the assembly of actin-rich structures or directly promote apical trafficking of Delta through Rab11-recycling endosomes.

A dynamic reorganization of the actin cytoskeleton into distinct cellular structures is also necessary to ensure successful myogenesis. Filopodial protrusions are crucial for the attachment of FCMs to the founder cell and growing myotube, and for the initiation of the fusion process (Ruiz-Gomez et al., 2000; Sens et al., 2010). The recognition and adhesion of myoblasts depends on members of the immunoglobulin superfamily (IgSF) that are expressed specifically in myoblasts in a ring-like structure. The interaction of these proteins leads to the formation of a cell communication structure, which has been termed fusion-restricted myogenic adhesive structure (FuRMAS) or podosome-like structure (Kesper et al., 2007; Richardson et al., 2007; Sens et al., 2010). The cytodomains of the IgSFs trigger the activation of WAVE in founder cells, and of WAVE and WASP in FCMs (Sens et al., 2010; Kaipa et al., 2013). In FCMs, WAVE- and WASP-mediated Arp2/3 activation results in the formation of a dense F-actin focus that accumulates at the interface of adhering myoblasts (Kesper et al., 2007; Sens et al., 2010). Electron microscopy studies have revealed that WASP is required for the formation of fusion pores at apposing myoblasts during embryonic and indirect flight muscle development (Massarwa et al., 2007; Berger et al., 2008, Dhanyasi et al., 2015). These fusion pores expand until full cytoplasmic continuity is achieved, and WASP has implicated to be required for fusion pore expansion (Massarwa et al., 2007). It has been discussed that WASP is required for the removal of membrane residuals during membrane vesiculation. WHAMY might contribute to this process, but the detailed mechanistic contribution of WHAMY in fusion pore formation needs to be addressed in future studies by ultrastructural analyses.

The following strains were used: w¹¹¹⁸, y¹ w^67c23; P{EPgy2}WhamyEY13506, y¹ w^67c23; P{GSV6}GS10268/TM3, hml^Δ-Gal4, hmlΔ-Gal4 EGFP, Lifeact–EGFP, neur-Gal4, ptc-Gal4, sca-Gal4, Df(3R)3450 (Bloomington Stock Center, Bloomington, IN) and wsp¹ (Ben-yaacov et al., 2001). The whamy mutant alleles whamy^Δ109, whamy^Δ65 and whamy^Δ58 were generated by imprecise transposon excision. The UAS-whamy transgenes were generated by ΦC31-integrase-mediated integration into the landing site M{3xP3-RFP.attP’}ZH-86FB^−/− as described previously (Stephan et al., 2011). Full-length wasp, whamy and whamy^ΔCRIB fragments were amplified by PCR and cloned into Gateway Entry Vectors (pENTR D-TOPO, Invitrogen). The inserts were sequenced and subcloned into pUASt-attB-rfA-mCherry, pUASt-attB-rfA-EGFP (Drosophila Genomics Resource Center) by LR in vitro recombination (Invitrogen) as previously described (Chen et al., 2014). All Drosophila strains were kept on standard medium and crosses were performed at 25°C unless otherwise indicated.

Drosophila S2R+ cells were propagated in 1× Schneider's Drosophila medium as described previously (Fricke et al., 2009). S2R+ cells were transfected as described previously (Kumar et al., 2009). For GST pulldown experiments, GST–Rac1 or GST–Cdc42 was expressed using E. coli Rosetta 2 (DE3) (Novagen). Lysis was performed by sonification in 1× PBS, 2 mM MgCl₂, 2 mM DTT, 10% glycerin and protease inhibitor cocktail (Roche). After centrifugation, GST proteins were loaded with GDP or GTPγS, as described previously (Fricke et al., 2009), and immobilized on glutathione resin according to the manufacturer's instructions (GE Healthcare). S2R+ cells transfected with pUAST-whamyEGFP, pUAST-whamyΔ^CRIB1EGFP, pUAST-whamyΔ^CRIB2EGFP or pUAST-whamyΔ^CRIB1+2EGFP and Act5cGal4 were harvested and lysed in 1 ml ice-cold lysis buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 1 mM DTT, 1.5 mM MgCl₂, 4 mM EDTA, 1% Triton X-100, 10% glycerin, protease inhibitor cocktail). After centrifugation, 500 μl of the cytoplasmic supernatant was added to 50 μl of loaded (GST–Rac1 or GST–Cdc42) glutathione resin. Following incubation for 1 h at room temperature, unbound proteins were washed away using four washes of 1 ml 1× PBS. 20% of the preparation was used per lane on standard SDS-PAGE and was analyzed by western blot analyses. Co-immunoprecipitations were performed as described previously (Bogdan et al., 2005).

GST-tagged WHAMY-FL, WASP and DiaC, as well as constitutively active Rac1-V12 and Cdc-V12 were expressed in E. coli strain ArcticExpress after induction with 0.7 mM IPTG at 12°C at an optical density >1.5. The cells were harvested after 20 h and the GST-tagged proteins were purified by affinity chromatography with glutathione-conjugated agarose (Sigma-Aldrich) using standard procedures. The GST tag was subsequently cleaved by PreScission protease (GE Healthcare) overnight, and the proteins were further purified by size exclusion chromatography on an Äkta Purifier System equipped with a HiLoad 26/60 Superdex S200 column (GE Healthcare) The WHAMY- and WASP-rich fractions were pooled, dialyzed against storage buffer (150 mM KCl, 1 mM DTT, 60% glycerol and 20 mM HEPES pH 7.4) and stored at −20°C. The expression and purification of recombinant Drosophila profilin was essentially performed as described previously (Breitsprecher et al., 2008). Arp2/3 complex was purified from bovine brain by affinity chromatography using a GST-tagged WCA domain from human WAVE1 coupled to glutathione-conjugated agarose essentially, as described previously (Loisel et al., 1999), and further purified by size exclusion chromatography on a HiLoad 26/60 Superdex S200 column. Rabbit skeletal muscle actin was extracted from acetone powder and purified according to Spudich and Watt, 1971. Fractions of G-actin were labeled with ATTO 488 maleimide (ATTO-TEC) or N-(1-Pyrene)maleimide (Life Technologies) at Cys³⁷⁴ according to the manufacturer's protocols.

For western blot analysis, adult guts of third-instar wandering larvae with the indicated genotype were homogenized in lysis buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 1.5 mM MgCl₂, 4 mM EDTA, 10% glycerol, 1% Triton X-100, protease inhibitor cocktail). Equal amounts of protein lysates were separated by SDS-PAGE (10%) and analyzed by western blotting. The affinity purified anti-WHAMY antibody was used 1:500. Anti-WHAMY antibodies were generated against Strep-tagged native WHAMY expressed in a cell-free wheat-germ lysate system (RTS 500, 5Prime). Rabbits were immunized with purified proteins (Davids Biotechnologie, Regensburg, Germany). Affinity purification was done as described previously (Zobel et al., 2015).

Actin polymerization assays were generally performed as described previously (Junemann et al., 2013). Polymerization of 2 µM rabbit skeletal muscle G-actin (5% pyrene-labeled) in 1× KMEI-polymerization buffer (50 mM KCl, 1 mM MgCl2, 1 mM EGTA, and 10 mM imidazole, pH 7.0) in the presence of various concentrations of WHAMY-FL and other proteins as indicated was monitored by a Synergy 4 fluorescence microplate reader (Biotek).

For each experimental condition, 150 µl of glutathione-conjugated agarose bead slurry were first washed three times with ice cold pulldown buffer containing 20 mM HEPES pH 7.4, 50 mM NaCl, 1 mM DTT, 5 mM MgCl₂, 2 mM benzamidine and 5% (v/v) glycerol. GST-tagged proteins were diluted in 1 ml pulldown buffer (final concentrations: 3 µM GST–WASP, 10 µM GST, 10 µM GST-Cdc42 and 10 µM GST-Rac1), then clear spun at maximum speed in a tabletop centrifuge for 15 min to remove precipitates. The supernatants of these protein solutions were then added to the washed beads and incubated for 1 h on a rotating wheel in the cold to allow binding. Subsequently the beads were spun for 2 min at 8000 g, the supernatants were discarded and the beads were incubated with 1 ml of the non-tagged protein solutions (2 µM WHAMY-FL, 10 µM Cdc42 and 10 µM Rac1 in pulldown buffer) for additional 2 h while slowly rotating. After four washing steps with pulldown buffer, bound proteins were eluted from the beads with SDS sample buffer and analyzed by SDS-PAGE.

WHAMY-FL and reference proteins from the gel filtration markers kit (Sigma-Aldrich) were diluted in filtration buffer [20 mM HEPES pH 7.4, 150 mM KCl, 1 mM DTT, 5% (v/v) glycerol] and analyzed by size-exclusion chromatography on a Superdex 200 Increase 10/300 GL gel filtration column using an ÄKTA Purifier system (GE Healthcare). Each run was performed three times. The calibration curve was obtained by plotting the decadic logarithm of the molecular mass [log(MW)] for each calibration protein against the ratio of elution volume (Ve) and void volume (Vo), which was 8.4 ml.

WHAMY-FL (100 nM final concentration) and profilin (5 µM final concentration) were pre-diluted in 1× TIRF buffer [20 mM imidazole pH 7.4, 50 mM KCl, 1 mM MgCl₂, 1 mM EGTA, 20 mM β-mercaptoethanol, 0.5 mM ATP, 15 mM glucose, 2.5 mg/ml methylcellulose (4000 cP), 20 µg/ml catalase, 100 µg/ml glucose oxidase]. The assays were started by adding G-actin (1 µM final concentration, 10% Atto488-labeled) and flushing these mixtures into the pre-coated flow chambers (Junemann et al., 2013). Time-lapse recordings were acquired using a Nikon Eclipse TI-E inverted microscope equipped with a TIRF Apo 60× objective and an Ixon3 897 EMCCD camera (Andor). Frames were captured every 2 s for 10 min with an exposure time of 50 ms. Two sectors per experiments were recorded using a motorized microscope stage. The pixel size corresponds to 0.27 µm. The nucleation efficacies were obtained by counting and averaging the number of nucleated actin filaments in an area of 80 µm×80 µm after 3 min at two positions in three independent experiments. The elongation rates of 25 filaments for each experimental condition were measured by manual tracking of growing barbed ends using ImageJ software. The lengths of five filaments for each experimental condition were measured manually every 10 s for a period of 300 s in absence of profilin and for 600 s in presence of profilin using the segmented line tool of ImageJ.

Pupal thorax epithelia of staged APF pupae were dissected in ice-cold PBS and fixed for 20 min in 4% paraformaldehyde in PBS. Epithelia were permeabilized in 0.15% Triton X-100 in PBS for 1 h, blocked for 1 h in 3% BSA and stained with primary antibody overnight at 4°C and then for 2 h at room temperature with the secondary antibody. Larval wing discs were dissected and stained in the same fashion. The following primary antibodies were used: mouse anti-Cut (1:10, DSHB), rat anti-Elav (1:10, DSHB), anti-Integrin βPS (1:50, DSHB), mouse anti-GFP (1:1000, Invitrogen), rabbit anti-mCherry (1:1000, Invitrogen). Secondary antibodies were Alexa-Fluor-488-, 568- or 647-conjugated anti-rabbit, anti-rat and anti-mouse IgG (1:1000, Invitrogen). F-actin was stained with Alexa-Fluor-647–phalloidin (1:100, Invitrogen). Samples were mounted in Vectashield (Vectorlabs) or Fluoromount (Southern Biotech). The rP298 enhancer trap line (Nose et al., 1998) was used to label muscle founder cells. The polyclonal anti-β3-Tubulin antibody (from rabbit, 1:12,000, Buttgereit et al., 1996) was used to analyze the somatic body wall musculature and a polyclonal anti-β-galactosidase (1:5000) antibody (Cappel) was used to identify balancer-carrying embryos and to visualize rP298–lacZ expression as described previously (Schafer et al., 2007).

SIM images were taken with an ELYRA S.1 Microscope (CellObserver SD, 63×/1.4 NA oil-immersion objective; Zeiss) with the software Zen 2010 D (Zeiss). Image acquisition was performed as described previously (Zobel and Bogdan, 2013). For isolation of macrophages, five white pupae or ten third-instar larvae were bled into 400 µl Shields and Sang M3 medium containing 10% FBS and 1× penicillin–streptomycin. The cell suspension was transferred onto a chambered cover glass and cells were incubated for 1 h at 25°C. Cell fixation and phalloidin staining were performed as previously described (Zobel and Bogdan, 2013).

Live imaging of macrophages in prepupae and in the pupal wing was performed with a CellObserver SD spinning disk microscope (Zeiss) as reported recently (Sander et al., 2013). The cuticles of staged pupae (16 h APF) were completely removed and pupae were placed on the lateral side and glued on glass bottom culture plates. Cells were imaged through a 40× Plan-Apochromat (NA 1.3) oil-immersion objective using an inverted Zeiss Axio Observer.Z1 microscope with a CCD camera (Axiocam MRm camera, 6.45 µm×6.45 µm) and the 488-nm laser line. For wounding assays, we used the UV laser ablation system DL-355/14 direct, from Rapp OptoElectronics. The ZEN software 2012 (Zeiss) was used for acquisition, and Zen as well as Fiji (ImageJ) software was used for processing of the images. For imaging living pupae or isolated macrophages, we used only 25% laser intensity of the 488-nm argon laser to avoid photodamage.

Prepupal macrophages were tracked manually using Fiji and the manual tracking plug-in. To quantify the migratory behavior of pupal macrophages we used the Histogram-based Macrophage Migration Score (HMMS) as described recently (Lammel et al., 2014).

We thank M. Sander for help during initial cloning of WHAMY constructs and cell transfection experiments. Meike Bechtold and Anna Julia Squarr for critical reading of the manuscript. We also thank the Bloomington Stock Center and VDRC for fly strains. We thank Jörg Grosshans for profilin and Diaphanous expression plasmids.